ClusterViSu, a method for clustering of protein complexes by Voronoi. tessellation in super-resolution microscopy

Transcription

1 ClusterViSu, a method for clustering of protein complexes by Voronoi tessellation in super-resolution microscopy Leonid Andronov 1,2,3,4, Igor Orlov 1,2,3,4, Yves Lutz 1,2,3,4, Jean-Luc Vonesch 1,2,3,4, Bruno P. Klaholz 1,2,3,4,* 1 Centre for Integrative Biology (CBI), Department of Integrated Structural Biology, IGBMC (Institute of Genetics and of Molecular and Cellular Biology), 1 rue Laurent Fries, Illkirch, France; 2 Centre National de la Recherche Scientifique (CNRS) UMR 7104, Illkirch, France; 3 Institut National de la Santé et de la Recherche Médicale (INSERM) U964, Illkirch, France; and 4 Université de Strasbourg, Strasbourg, France. *To whom correspondence should be addressed. klaholz@igbmc.fr

2 Figure S1 Evaluation of the quality of Voronoi image reconstruction (A-C) Simulated images of two lines separated by 40 nm and composed of localization events with a standard deviation of 10 nm in histogram (A), Gaussian (B) and Voronoi-based interpolated (C) local density representations (same as Figs. 1A-C, but with line indicating the line profile). (D) One-pixel width profiles along the colored lines of images A-C. (E-G) FRC curves calculated from an image region in Fig. 1E in histogram (blue), Gaussian (green) and Voronoi local density (red) representations; pixel size of the image: 5 nm (E), 3 nm (F), 2nm (G). The graphs show only the frequency range from 0 to 0.05 nm -1 for consistency with Fig. 1F. Suppl. Fig. S1

3 Figure S2 Simulations of different co-localization scenarios (A) Voronoi diagrams of simulated double-color clusters. (B) Results of segmentation of the double-color clusters in ClusterViSu with the overlay area shown in yellow and the calculated colocalization values shown below. rig and rir are the Cartesian coordinates of the centers of the green and the red clusters, respectively. The simulated radius of clusters is 50 nm; density of events in clusters = nm -2, in background = nm -2. Localizations in the clusters and in the background are distributed randomly and independently for each color. Suppl. Fig. S2

4 Figure S3 Effect of the localization error on Voronoi clustering (A) Results of segmentation of a simulated cluster data with different amounts of localization error. σloc is the standard deviation of the normal distribution used for simulating the coordinates of the points. Clusters containing 5 or less molecules were removed. (B) Complete segmented FOVs of the simulated data with high localization errors σloc demonstrating remaining clusters at σloc = 60 nm and σloc = 100 nm. (C) The Voronoi diagrams of the upper-left corner of the simulated data with different localization errors. The density of clustered localizations is 0.01 nm -2, the density of background is nm -2. Suppl. Fig. S3

5 Figure S4 Effect of multiple localizations on Voronoi clustering (A&B) Results of segmentation of simulated cluster data with different numbers of relocalizations. Clusters containing 5 or less molecules were removed. (C) Comparison of Voronoi diagrams with different numbers of re-localizations (n). The density of clustered localizations (at n = 1) is nm -2 ), the density of background is 20% of the cluster density; the localization precision σloc = 10 nm. Suppl. Fig. S4

6 Figure S5 Effect of multiple localizations on Monte-Carlo simulations (A&B) Voronoi polygon distributions of spatially random datasets with different numbers of relocalizations (n) and results of Monte-Carlo simulations taking into account multiple localizations (nsim). (C) Simulated clustered datasets with re-localizations (n = 5) segmented by thresholds determined by simulations without (nsim = 1) and with (nsim = 5) multiple localizations. The density of localizations (at n = 1) is nm -2 (A&B); the density of clustered localizations (at n = 1) is nm -2, the density of background is 20% of the cluster density (C); the localization precision σloc = 10 nm (A-C). Suppl. Fig. S5

7 Figure S6 Comparison of ClusterViSu with SR-Tesseler. (A&B) Clustered dataset segmented in ClusterViSu with threshold determined by Monte-Carlo simulations (A) and in SR-Tesseler with threshold determined by average localization density 17 δ 1 i = 2 <δ> (B). Density of clustered localization is nm -2, background density is 0.01 nm -2. Clusters containing five or less molecules were removed. (C) Retrieval of the equivalent radius of clusters in ClusterViSu and SR-Tesseler. Clusters containing two or less molecules were removed. (D&E) Number of clusters detected by ClusterViSu (red curves) and SR-Tesseler (blue curves) as function of background density. Nmin, minimal number of events in clusters kept for quantification. (F) Variation of segmentation threshold determined by Monte-Carlo simulations in ClusterViSu (red curve) or by average localization density in SR-Tesseler 17 (cyan and blue curves), depending on background density. Suppl. Fig. S6